best xml parser to use

[ftumsh]

Lo all,
Given that
1) the xml I have to use is quite simple (no cdata or PI etc)
2) but possibly large (< 5meg)
3) I need to know a tag before it's children are parsed
4) have a small memory footprint
5) be very fast
6) Linux only

Should I be using XML::Parser or XML::LibXML?

Thx
John

[Tanktalus]

My general decision tree goes sorta like this (non-XML parts removed):

 /----------\           +-----------+
< Parse XML? >  --NO>-- | (removed) |
 \----------/           +-----------+
      |
     YES
      V
      |
+---------------+
| Use XML::Twig |
+---------------+

Hopefully this decision tree helps you decide what is best.

;-)

[ftumsh]

heh :)
I normally always use xml::twig, but I don't actually know the format of xml so I can't in this case. I've been googling and I'm going to try xml::sax.

I appear to have only got xml::libxml::sax, is this good enough? the pod mention it might not be any good for production use. What others could I use?

[rvosa]

What do you mean by "I don't actually know the format of xml"? How will switching parsers (though, technically, XML::Twig is a front-end, not a parser itself) fix that?

[Jenda]

if ($xml->is_simple() and $xml->is_table_like()) {
  use XML::RAX;

} elsif (size($xml) < too_big()) {
  use XML::Simple;

} else {
  use XML::Twig;
  ...
  my $Data = $DataObj->simplify(forcearray => [...], keyattr =>{ ... }
+, group_tags => {...});
}

Update 2007-2-6: Tastes change. While I would probably still use XML::RAX for some tasks with table-like XML and XML::Simple for very simple XMLs, I'd most probably use my XML::Rules now. May look a bit twisted at first, but it's convenient and powerfull. IMHO of course ;-)

[coreolyn]

It seems to me that performance is so much better just parsing xml with regex's I quit caring that it's xml. I might add some coding time to my development but I really don't see the overall benifit to xml modules unless I have to provide xml output -- even then it's 'iffy'.

[planetscape]

What Tanktalus++ said. (Another testimonial.)

HTH,

[GrandFather]

Agreed, although I sometimes still use XML::TreeBuilder.

Love Tanktalus's decision tree though. :)

[xdg]

I'm not sure how it stacks up against these criteria, but if you're not 100% sure that your incoming data is entirely valid, you might want to check out XML::Liberal (a standin for LibXML). Avoiding re-parsing a large XML file because of a small nit might be worthwhile form of efficiency.

The author gave a nice Lightning Talk about it at YAPC::NA.